Michael Connelly's Proving Ground

An AI technical and ethical tutorial disguised as a legal thriller

I read mysteries. Lots of mysteries. The combination of escape and learning keeps me coming back

In order to write a great mystery or legal thriller, the author has to focus on a deep understanding of a specific subject. Whether it’s the heart of the case, complex conspiracies, how the poison works, steps in an investigation, the meaning of clues, or any of the other major plot devices, the author has to understand the topic on which the mystery turns.

That means that you can learn just enough to be dangerous by reading the latest pulp fiction on the bestseller list. Often, the author’s research and explanation are good enough for a detailed overview of the subject.

I’ve been writing about and explaining complex technical topics (like AI, coding, systems theory) for a long time. The current crush of opinion and hype about AI coupled with its rapid evolution make simplification hard. Simplification always leaves meaningful detail out of the description.

Learning about a new technology means getting better and better at understanding what you don’t know and learning how to fill the gaps. It takes some time to learn how to navigate around the rabbit holes. You can go pretty deep in tecnical subjects. Not all of it is useful for understanding and decision making by someone who is not a builder of product.

That brings me to the new (ish) novel by Michael Connelly, Proving Ground. Connelly is the author of the Harry Bosch and Lincoln Lawyer novels. He’s prolific and manages to avoid being overly formulaic in his work. The characters are familiar, the mysteries are unique.

Proving Ground is the story of a young man who killed his girlfriend because an LLM told him to. The plot covers a broad array of AI topics. From product liability and ethics to the inner workings of conversational AI, the legal case depends on being able to simply tell the AI story. Trial prep depends on having an easy to understand narrative.

As the story unfolds, the reader is taken on a journey through complex technical, legal and ethical topics. You learn how an LLM conversation can evolve, what’s happening in the background, why guardrails fail, and how fault is determined when something goes wrong. It’s particularly juicy becasue the bad guys are ‘evil’ tech company executives with delusions of grandeur. (I imagine this is the early stage of a flood of fictional work where the villains are billionaires and billionaire wannabes.)

If you want a firmer grasp of the underlying operations of AI tools, this is like a spoonful of sugar for the medicine. As the plot unfolds, the tutorial gets deepet. But, since its hung on a mystery narrative, it is easier to consume the technical information.

It’s a mystery. The good guys win. I’ve never read a mystery where the good guys don’t win. But, along the way, you get a good, detailed overview of the inner workings of AI.

=========================
Photo by Growtika